Autonomous driving system and method of controlling same

The present application claims priority to Korean Patent Application No. 10-2022-0067025, filed May 31, 2022, the entire contents of which are incorporated herein for all purposes by this reference.

The present disclosure relates to an autonomous driving system that tiers learning data according to the driving environment of an autonomous vehicle and a method of controlling the same.

Generally, an autonomous driving system may obtain the greatest amount of raw data as possible and normalize learning data by learning of the obtained raw data in order to improve the accuracy of an autonomous driving algorithm.

Meanwhile, a new factor for the driving environment of the autonomous vehicle is applied to the autonomous driving algorithm, the autonomous driving system may additionally obtain raw data for new driving environment and improve learning data using the additional raw data.

However, in the above-described method, whenever a new factor for the driving environment is applied to the autonomous driving algorithm, it is required to additionally obtain raw data for the new driving environment, and existing learning data disappears. Thus, there is a problem in that it is difficult to examine the history of the learning data.

In addition, in a data tree having a hierarchical structure to which a variety of variables are added, the learning data is only transferred to lower data layers. However, in general, learning results of lower layers are reflected on higher layers.

The foregoing is intended merely to aid in the understanding of the background of the present disclosure, and is not intended to mean that the present disclosure falls within the purview of the related art that is already known to those skilled in the art.

Accordingly, the present disclosure has been made keeping in mind the above problems occurring in the related art, and the present disclosure is intended to examine a history of learning data according to the driving environment by tiering the learning data according to the driving environment of an autonomous vehicle.

The present disclosure is also intended to improve the accuracy of an autonomous driving algorithm by performing update between root learning data and layer learning data in a bidirectional manner when tiering the learning data.

The objective of the present disclosure is not limited to the aforementioned description, and other objectives not explicitly disclosed herein will be clearly understood by those skilled in in the art from the description provided hereinafter.

In order to achieve the above objective, according to one aspect of the present disclosure, there is provided a method of controlling an autonomous driving system. The method may include: generating root learning data by performing learning for raw data; generating a plurality of first layer learning data by performing learning, to which driving environment variables of an autonomous vehicle are applied, for the root learning data; and updating the root learning data from the plurality of first layer learning data depending on whether or not an integration condition of the plurality of first layer learning data is met.

Also provided is a method of controlling an autonomous driving system. The method may include: generating first learning data; generating at least one piece of second learning data corresponding to a lower layer of the first learning data by performing learning, to which driving environment variables of an autonomous vehicle are applied, for the first learning data; and updating the first learning data from the at least one piece of second learning data depending on whether or not the at least one piece of second learning data meets a predetermined first condition.

Also provided is an autonomous driving system including: a learning device generating root learning data by performing learning for raw data and generating a plurality of first layer learning data by performing learning, to which driving environment variables of an autonomous vehicle are applied, for the root learning data; and a learning control device controlling the learning performed by the learning device and updating the root learning data from the plurality of layer learning data depending on whether or not an integration condition for the plurality of layer learning data is met.

According to the present disclosure, even in the case that new factors for learning data are applied to an autonomous driving algorithm, it is possible to examine a history of learning data according to the driving environment by tiering the learning data according to the driving environment of an autonomous vehicle.

In addition, according to the present disclosure, it is possible to improve the accuracy of an autonomous driving algorithm by performing update between root learning data and layer learning data in a bidirectional manner when tiering the learning data.

The effects of the present disclosure are not limited to the aforementioned effects, and other effects not explicitly disclosed herein will be clearly understood by those skilled in in the art from the description provided hereinafter.

Hereinafter, embodiments disclosed in the present disclosure will be described in detail with reference to the accompanying drawings, in which identical or similar constituent elements are given the same reference numerals regardless of the reference numerals of the drawings, and a repeated description thereof will be omitted.

In the description of the present disclosure, when it is determined that the detailed description of the related art would obscure the gist of the present disclosure, the detailed description thereof will be omitted. In addition, the attached drawings are merely intended to be able to readily understand the embodiments disclosed herein, and thus the technical idea disclosed herein is not limited by the attached drawings, and it should be understood to include all changes, equivalents, and substitutions included in the idea and technical scope of the present disclosure.

It will be understood that, although the terms “first”, “second”, etc., may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another element.

As used herein, the singular forms “a”, “an”, and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise.

It will be further understood that the terms “comprise”, “include”, “have”, etc., when used in this specification, specify the presence of stated features, integers, steps, operations, elements, components, and/or combinations thereof but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or combinations thereof.

In addition, the term “unit” or “control unit” included in names is merely a term used in naming a controller controlling specific functions of a system but should not be interpreted as a generic function unit.

is a diagram illustrating an example configuration of an autonomous driving systemaccording to an embodiment of the present disclosure.

Referring to, the autonomous driving systemmay include a raw data storage device, a driving environment information recognizing device, a learning control device, a learning device, a learning data storage device, and a learning data output device.

According to an exemplary embodiment of the present disclosure, the autonomous driving systemmay include a processor (e.g., computer, microprocessor, CPU, ASIC, circuitry, logic circuits, etc.) and an associated non-transitory memory storing software instructions which, when executed by the processor, provides the functionalities of, for example, the learning control deviceand the learning device. Herein, the memory and the processor may be implemented as separate semiconductor circuits. Alternatively, the memory and the processor may be implemented as a single integrated semiconductor circuit. The processor may embody one or more processor(s).

The raw data storage devicemay receive raw data from sensors of an autonomous vehicle or sensors disposed inside traffic infrastructure (e.g., traffic lights and road signs) and store the received data. The sensors provided in the autonomous vehicle and the traffic infrastructure may be respectively implemented as an acoustic sensor, a light sensor, an electromagnetic sensor, or the like, in the form of a radar sensor, a light detection and ranging (LiDAR) sensor, a camera, a microphone, an accelerometer, a gyroscope.

The driving environment information recognizing devicemay receive information regarding the driving environment of the autonomous vehicle and transfer the received information to the learning control device. Here, the driving environment of the autonomous vehicle refers to an environment related to a natural condition or a social situation directly or indirectly affecting the driving of the autonomous vehicle. For example, factors of the driving environment of the autonomous vehicle may include a country where the autonomous vehicle is driving, as well as a traffic system and an area of the country. Meanwhile, the driving environment information recognizing devicemay recognize the driving environment of the autonomous vehicle by exchanging information with the raw data storage device.

The learning control devicemay control learning performed by the learning deviceon the basis of information regarding the driving environment, raw data, and learning data stored in the learning data storage device. In the present embodiment, the learning data may include root learning data and layer learning data.

The learning devicemay generate the root learning data by performing learning for the raw data by the learning control deviceand generate first to Nth layer learning data by performing the learning, to which driving environment variables of the autonomous vehicle are applied, for the root learning data (where ‘N’ is a natural number equal to or greater than 2). In the present embodiment, the driving environment variables of the autonomous vehicle may be respectively expressed in different levels.

More specifically, the learning devicemay generate a plurality of first layer learning data corresponding to a lower layer of the root learning data by performing the learning, to which the driving environment variable of the autonomous vehicle corresponding to a first level are applied, for the root learning data. Afterwards, the learning devicemay generate a plurality of Nth layer learning data corresponding to the lower layer of the (N−1)th layer learning data by performing the learning, to which the driving environment variable of the autonomous vehicle corresponding to the Nth level are applied, for at least one of the plurality of (N−1)th layer learning data. That is, the learning devicemay tier the root learning data and first to Nth layer learning data into a tree structure. Thus, the autonomous driving systemmay examine the history of the learning data according to the driving environment even in the case that new factors for the driving environment are applied to the autonomous driving algorithm.

Meanwhile, the learning devicemay perform the learning for data input according to a machine learning algorithm. The machine learning algorithm may be implemented as at least one selected among a supervised learning algorithm, an unsupervised learning algorithm, a reinforcement learning algorithm, and combinations thereof.

The learning control devicemay update the root learning data from the plurality of layer learning data depending on whether or not an integration condition for the plurality of layer learning data is met. Here, the integration conditions may be met when the plurality of layer learning data, generated by the learning to which the driving environment variables are applied, include common learning information. In addition, the learning control devicemay update the first to Nth layer learning data from the root learning data depending on whether or not a propagation condition for the root learning data is met. Here, the propagation condition may be met on the basis of the accuracy of the autonomous driving algorithm according to the root learning data. That is, the learning control devicemay update the root learning data and the layer learning data in a bidirectional manner in order to increase the accuracy of the autonomous driving algorithm. The operation of updating the learning data by the learning control devicewill be described more specifically later with reference to.

The learning data storage devicemay store the root learning data and the layer learning data generated by the learning device.

The learning data output devicemay receive the root learning data and the layer learning data from the learning control deviceand output the root learning data and the layer learning data to a driving controller of the autonomous vehicle according to the driving environment of the autonomous vehicle.

is a diagram illustrating an example process of learning data tiering by the learning deviceillustrated in.

Referring to, the learning devicemay generate root learning data by performing learning for raw data.

Afterwards, the learning devicemay tier the learning data by sequentially generating first layer learning data A, B, and C, second layer learning data D, E, F, G, and H, and third layer learning data I, J, K, L, M, and N by performing learning, to which the driving environment variables of the autonomous vehicle are applied, for the root learning data.

More specifically, the first layer learning data A, B, and C may be generated by performing learning, to which the driving environment variable corresponding to a first level LEVEL 1 is applied, for the root learning data, whereas the second layer learning data D, E, F, G, and H may be generated by performing learning, to which the driving environment variable corresponding to a second level LEVEL 2 is applied, for the first layer learning data A, B, and C. In addition, the third layer learning data I, J, K, L, M, and N may be generated by performing learning, to which the driving environment variable corresponding to a third level LEVEL 3 is applied, for the second layer learning data D, E, F, G, and H.

is a diagram illustrating an example variable according to the driving environment of an autonomous vehicle.

Referring to, the driving environment variable corresponding to a first level LEVEL 1 may be determined according to the traffic system of a country (e.g., a country where the steering wheel is provided on the left side, a country where the steering wheel is provided on the right side, and a country where the steering wheel is provided at the center) where the autonomous vehicle is driving. In addition, the driving environment variable corresponding to a second level LEVEL 2 may be determined according to the traffic system of a country (e.g., the Republic of Korea and the USA among countries where the steering wheel is provided on the left side and Japan and the United Kingdom among countries where the steering wheel is provided on the right side) where the autonomous vehicle is driving. Furthermore, the driving environment variable corresponding to a third level LEVEL 3 may be determined according to the area (e.g., the western, middle, and eastern areas of the USA, the Tokyo Metropolis and Hokkaido of Japan, and Scotland of the United Kingdom) of a country where the autonomous vehicle is driving.

is a diagram illustrating an example process in which the learning control deviceillustrated inupdates root learning data from layer learning data.

Referring to, third layer learning data L and M regarding the Tokyo Metropolis and Hokkaido of Japan have second layer learning data F regarding Japan as a parent node, and thus are in a sibling relationship. When the third layer learning data L and M in the sibling relationship meet an integration condition, the learning control devicemay update the second layer learning data F from the third layer learning data L and M so as to improve the accuracy of the autonomous driving algorithm regarding Japan.

In the same manner, when second layer learning data F and G in the sibling relationship meet an integration condition, the learning control devicemay update the first layer learning data B so as to increase the accuracy of the autonomous driving algorithm regarding the country where the steering wheel is provided on the right side.

Finally, when the first layer learning data A, B, and C in the sibling relationship meet an integration condition, the learning control devicemay update the root learning data from the first layer learning data A, B, and C so as to increase the accuracy of the autonomous driving algorithm according to the root learning data.

is a diagram illustrating an example process in which the learning control deviceillustrated inupdates layer learning data from root learning data.

Referring to, when the root learning data meets a propagation condition, the learning control devicemay sequentially update the first layer learning data A, B, and C, the second layer learning data D, E, F, G, and H, and the third layer learning data I, J, K, L, M, and N from the root learning data. Thus, the autonomous driving systemmay improve the accuracy of the algorithm according to the driving situation of the autonomous vehicle.

is a diagram illustrating an example process in which the autonomous driving system applies the driving environment variable when the driving environment of the autonomous vehicle is changed.

Referring to, when the autonomous vehicle moves from Scotland of the United Kingdom to the eastern area of the USA, the driving environment information recognizing devicemay transfer information regarding the driving environment to the learning control device, and the learning control devicemay control the learning deviceto perform learning to which the driving environment variable regarding the western area of the USA are applied. In addition, the learning data output devicemay output learning data regarding the western area of the USA to the driving controller.

Although the above description has been provided with reference toby assuming a situation in which the driving environment variables are categorized according to the traffic system and geographical division, this is for illustrative purposes only. It will be apparent to those skilled in the art that a variety of driving environment variables affecting the autonomous driving may be used.

is a flowchart illustrating a method of controlling the autonomous driving system according to an embodiment of the present disclosure.